Modelling bidding strategies of flexibilities under uncertain price forecasts - An agent-based modelling approach

Abstract

Motivation Flexibilities such as battery storages, pumped-hydro plants, and demand-response are likely to contribute significantly to future energy systems with a high share of variable renewable energy sources (RES). The interplay of these technologies on energy markets might determine future electricity prices. Thus, a detailed understanding of bidding strategies is required in order to model future energy markets and their dynamics - a key component for energy-related investment decisions. Modelling the competition of flexibilities, however, requires adequate simulation of uncertainties with regard to electricity price forecasts (see, e.g., [1]) and strategies to cope with these. We present bidding strategies for flexibilities in a competitive environment that consider uncertainty of price forecasts and competitors' behaviour. Method We simulate the day-ahead market dynamics using the agent-based model (ABM) for electricity markets AMIRIS [2]. This allows representing various actors and their specifics such as their objectives, risk aversion or affinity, different levels of information, and interactions with the market environment [3]. Based on previous work for the assessment of storage technologies with AMIRIS [4-6], we distinguish two different modelling approaches: First, actors with large capacities with significant market power and thus a respective feedback of their actions on the market, and second, a multitude of actors with smaller capacities who need to consider their competitors' impact on market prices. Therefore, in project ERAFlex II , we implement two bidding strategies for actors with market power and perfect knowledge (named "MinCost" & "MaxProfit"), as well as two further bidding strategies for flexibilities with less market power and less information (named "PriceMedian" and "ResidualFeedback"). "MinCost" aims to minimize system cost, while "MaxProfit" maximizes individual profit considering market power. These strategies, which are similar to those used in optimization models, provide benchmarks for evaluating the performance of the other strategies. "PriceMedian" uses price forecasts to determine optimal charging and discharging times, while "ResidualFeedback" estimates the impact of its own and competitors' bids on final market prices using an estimated price-over-residual-load curve. Results In preliminary analyis, we show total profits of the storage operator(s), their energy discharge total, and total system costs for the four strategies in one year in a scenario for Germany with 80% RES share. The "MinCost" strategy has the lowest system cost but yields only about half the profits of the "MaxProfit" strategy. With 10 competing storage agents, the "PriceMedian" strategy is too simplistic, leading to limited profits and higher system costs. However, the "ResidualFeedback" strategy performs similarly well as the "MaxProfit" strategy - even with competing agents. Up to this point, our evaluation has focused on the competition among similar flexibility technologies in energy markets. However, future energy systems are expected to incorporate a variety of technologies with differing constraints, availabilities, and energy-to-power ratios. Modelling such systems could benefit from a more centralized approach that leverages machine learning to generate price forecasts and utilizes stochastic optimization to identify suitable bidding strategies . References [1] S. Beltrán, A. Castro, I. Irizar, G. Naveran, and I. Yeregui, "Framework for collaborative intelligence in forecasting day-ahead electricity price," Applied Energy, vol. 306, p. 118049, 2022. [2] C. Schimeczek et al., "AMIRIS: Agent-based Market model for the Investigation of Renewable and Integrated energy Systems," JOSS, vol. 8, no. 84, p. 5041, 2023, doi: 10.21105/joss.05041. [3] O. Kraan, G. J. Kramer, and I. Nikolic, "Investment in the future electricity system-An agent-based modelling approach," Energy, vol. 151, pp. 569-580, 2018. [4] L. Torralba Diaz et al., Effektive Rahmenbedingungen für einen kostenoptimalen EE-Ausbau mit komplementären dezentralen Flexibilitätsoptionen im Elektrizitätssektor - ERAFlex: Projektbericht, 2019. [Online]. Available: https://elib.dlr.de/132823/1/Endbericht.pdf [5] F. Nitsch, M. Deissenroth-Uhrig, C. Schimeczek, and V. Bertsch, "Economic evaluation of battery storage systems bidding on day-ahead and automatic frequency restoration reserves markets," Applied Energy, vol. 298, p. 117267, 2021, doi: 10.1016/j.apenergy.2021.117267. [6] F. Nitsch and M. Wetzel, "Profitability of Power-to-Heat-to-Power Storages in Scenarios With High Shares of Renewable Energy," in Energy Proceedings: Closing Carbon Cycles.